Trying to avoid using the ZDCs, Alán found an alternative SC&GL calibration:

SC = (1.0 ± 0.03465)*(2.794e-07*(bbcx-( 262.3 ± 629.6))+(3.438e-07*(bbcbb)))
GL = 6.57 +/- 0.62

I processed (in DEV) up to 2000 events in a few st_centralpro files scattered through fill 16852 (which was highlighted previously in my look at Run 12 UU BeamLine) and found...

1. BeamLine still sees a very similar luminosity dependence. Here is the P12ic (left) and above calibration (right) compared via BeamLine x0 (cutting on rank > -3 and trigger ID for the protected central trigger) for identical files, showing only a very small change:
   
2. Signed DCAs (of global partners of primary tracks from the highest ranked vertex with more than 25 hits) show a luminosity dependence, seen versus time (left) and zdcx (right) with the P12ic in red and the above calibration in blue, showing a similar dependence, but actually larger sDCA magnitudes:
   
   Perhaps an interesting result: a linear fit to the right plots give an intercept which is closer to (0,0) for the blue than the red, which is perhaps a point in favor of the above calibration.

Of note, I can see a slightly sinusoidal change in <sDCA> vs. φ for the two datasets, but this is almost certainly due to the slight shift of the BeamLine noted already.

-Gene